This issue is more a reminder and tracking issue, since I did not manage to reproduce it reliably yet.
Sometimes, in high-throughput scenarios, the coordinator the pub/sub system silently fails. Resetting the bagfile does not help, only restarting the coordinator does.